Find a Java library for Bayesian networks that might work

Introduction

I needed to run a Bayesian network in Java, so I searched for a nice library.

What i did

As a condition

--Maintenance is still ongoing --Distributed as a jar file --Complete tutorial

• Open Source

In the Bayesian network library I searched for a person who could learn the network structure and conditional probabilities.

result

The following two can be used

• weka
• mahout

Other than that, the following

--Netica-j (Error could not be resolved) --jayes (Build failed, GitHub update stopped 5-6 years ago and there is an unmaintained atmosphere) --BayesServer (It feels really good, but it costs about 70,000 yen)

Use of the library

weka Weka is a machine learning software developed at the University of Waikato in New Zealand, and jar files can also be used as GUI applications. This time, I will use the class in the jar file in Java. I used this for the data for learning. https://gist.github.com/carl0967/7a3588cd6f0d40d02a26

Below source code

import java.io.*;
import java.util.*;

import weka.core.converters.ArffLoader;
import weka.classifiers.Evaluation;
import weka.classifiers.bayes.BayesNet;
import weka.classifiers.bayes.net.search.SearchAlgorithm;
import weka.classifiers.bayes.net.search.local.SimulatedAnnealing;
import weka.classifiers.bayes.net.search.local.K2;
import weka.core.Instances;
import weka.core.converters.ConverterUtils.DataSource;

public class bayesNet{

	Instances instances;
	BayesNet bnet;
	Evaluation evaluation;

	public bayesNet(){}

	void setFile(File dataFile){
		try {
			ArffLoader al = new ArffLoader();
			al.setFile(dataFile);
			instances =  al.getDataSet();
			instances.setClassIndex(instances.numAttributes() - 1);
		} catch (Exception e) {
			e.printStackTrace();
		}

	}

	void buildClassifier(){
		try{
			bnet = new BayesNet();
			//Search algorithm generation
			SearchAlgorithm searchAlgorithm = new K2();
			//Set on BayesNet
			bnet.setSearchAlgorithm(searchAlgorithm);
			//Start classification
			bnet.buildClassifier(instances);
		} catch(Exception e){
			e.printStackTrace();
		}
		
	}

	void evalute(){
		try{
			//Evaluation
			evaluation = new Evaluation(instances);
			evaluation.evaluateModel(bnet, instances);
		} catch(Exception e){
			e.printStackTrace();
		}
	}

	void showResult(){
		System.out.println(evaluation.toSummaryString("Results\n",false));
	}

	public static void main(String[] args) {
		bayesNet classifier = new bayesNet();

		classifier.setFile(new File(args[0]));
		classifier.buildClassifier();
		classifier.evalute();
		classifier.showResult();
	}

}

Reference: Create a Bayesian network using Weka's API in Java

result

WARNING: An illegal reflective access operation has occurred
WARNING: Illegal reflective access by weka.core.WekaPackageClassLoaderManager (file:/Users/Mnb0130/unirvFILE/semi/Engineering special exercise/bayesNet/weka/weka.jar) to method java.lang.ClassLoader.defineClass(java.lang.String,byte[],int,int,java.security.ProtectionDomain)
WARNING: Please consider reporting this to the maintainers of weka.core.WekaPackageClassLoaderManager
WARNING: Use --illegal-access=warn to enable warnings of further illegal reflective access operations
WARNING: All illegal access operations will be denied in a future release
Results

Correctly Classified Instances          34               89.4737 %
Incorrectly Classified Instances         4               10.5263 %
Kappa statistic                          0.6122
Mean absolute error                      0.204
Root mean squared error                  0.299
Relative absolute error                 59.8749 %
Root relative squared error             73.3056 %
Total Number of Instances               38

If the correct answer rate is about 89% with a small data set, I think that it is an assumed movement operation. After all, I felt that using weka was appropriate in most cases.